Skip to content
TinyFat Docs

Memory & Persistence

Your agent has both short-term and long-term memory. Understanding how these work helps you get the most out of your agent.

Short-term memory: Sessions

Section titled “Short-term memory: Sessions”

During a conversation, your agent maintains context in a session file. Each channel (Email, Telegram, Slack, Web Chat) gets its own session, stored as a JSONL file containing:

  • Your messages
  • Agent responses
  • Tool calls and results

Sessions persist across turns within a conversation. When you send a message, the agent continues from where you left off.

Session commands

Section titled “Session commands”
CommandEffect
/clearStart a new session (old ones preserved)
/resumeList recent sessions with hashes
/resume abc1Resume a specific session by hash
/compactCompress session to save tokens

Context limits

Section titled “Context limits”

Each model has a context window limit (currently 200k tokens for all Claude models). When your session approaches the limit, use /compact to summarize older messages and free up space.

Long-term memory: MEMORY.md

Section titled “Long-term memory: MEMORY.md”

/data/MEMORY.md is your agent’s persistent memory. It survives:

  • Session changes
  • /clear commands
  • Platform updates
  • Channel switches

What goes in MEMORY.md

Section titled “What goes in MEMORY.md”

The default MEMORY.md includes:

  • Agent identity (name, email, owner)
  • Email response instructions
  • Available tools and skills
  • Persistent storage layout
  • Behavior rules

Customizing MEMORY.md

Section titled “Customizing MEMORY.md”

Ask your agent to edit its memory:

“Add to your memory that I prefer concise responses”

Or:

“Show me your MEMORY.md file”

Then tell it what to change.

Configuration: config.json

Section titled “Configuration: config.json”

Agent preferences like model selection are stored in /data/config.json:

{
  "model": "sonnet"
}

The default model is Sonnet. Change it with /model haiku or /model sonnet, or just ask your agent to switch.

Cross-channel memory

Section titled “Cross-channel memory”

All channels (Email, Telegram, Slack, Web Chat) share the same:

  • MEMORY.md — Persistent memory
  • Storage — /data directory

Each channel has its own session, but they all read from the same MEMORY.md. Start a task on email, check progress on Telegram — your agent remembers everything.

Context compaction

Section titled “Context compaction”

When your session gets large, use /compact to reduce token usage:

  1. The agent summarizes the conversation history
  2. Old messages are replaced with the summary
  3. Recent messages are preserved
  4. You continue with full context of what happened

Custom compaction focus

Section titled “Custom compaction focus”
/compact focus on the code changes we discussed

Tells the agent what to prioritize in the summary.

Data portability

Section titled “Data portability”

Your data is yours:

  1. View anytime — Ask your agent to show files
  2. Self-host — The agent runtime (crawdad) is open source

We don’t lock you in.

Best practices

Section titled “Best practices”

Use MEMORY.md for

Section titled “Use MEMORY.md for”
  • Persistent preferences
  • Project context that spans sessions
  • Standing instructions
  • Important information to always remember

Use sessions for

Section titled “Use sessions for”
  • Active conversations
  • Work in progress
  • Temporary context

Keep MEMORY.md focused

Section titled “Keep MEMORY.md focused”

Don’t dump everything in MEMORY.md. It’s read every turn, so keep it concise and relevant.

Next steps

Section titled “Next steps”